Paper: arXiv 2504.16892
Authors: John Armstrong, James Dalby, Rohan Hobbs
Abstract
We present an architecture for managing Collective Defined Contribution (CDC) schemes. The current approach to UK CDC can be described as shared-indexation, where the nominal benefit of every member in a scheme receives the same level of indexation each year. The design of such schemes rely on the use of approximate discounting methodologies to value liabilities, and this leads to intergenerational cross-subsidies which can be large and unpredictable. We present an alternative approach which we call Collective-Drawdown CDC. This approach does not result in intergenerational cross-subsidies since all pooling is performed by explicit insurance contracts. It is therefore completely fair. Moreover, this scheme results in better pension outcomes when compared to shared-indexation CDC under the same model parameters.
Complexity vs Empirical Score
- Math Complexity: 8.0/10
- Empirical Rigor: 3.0/10
- Quadrant: Lab Rats — high math complexity, limited empirical testing
Why this score: The paper involves advanced mathematical modeling of pension scheme liabilities, fairness conditions, and optimization, including references to stochastic calculus and neural network-based optimal control, but it presents a theoretical architecture and conceptual comparison rather than empirical backtests or real-world implementation details.
Research Flowchart
flowchart TD
A["Research Goal<br>Design CDC scheme<br>without intergenerational<br>cross-subsidies"] --> B{"Methodology"}
B --> C["Analyze Shared-Indexation CDC<br>Identify valuation flaws"]
B --> D["Propose Collective-Drawdown CDC<br>Implement explicit insurance"]
C --> E["Simulate Liability Valuation<br>Approximate discounting"]
D --> F["Simulate Liability Valuation<br>Insurance contract pricing"]
E --> G["Computational Analysis<br>Measure cross-subsidies & outcomes"]
F --> G
G --> H{"Key Findings"}
H --> I["Collective-Drawdown eliminates<br>intergenerational cross-subsidies"]
H --> J["Collective-Drawdown yields<br>superior pension outcomes<br>vs Shared-Indexation"]
Evidence Card
What the authors claim, on what data, and who has checked it. Read from the paper's own text by a language model: every quote is verbatim, nothing was re-run. How to read this card
Proposes
The paper proposes a Collective‑Drawdown CDC architecture that replaces shared‑indexation and uses explicit insurance contracts to pool risk, claiming it is fair and yields better pension outcomes.
Source sentence
We present an architecture for managing Collective Defined Contribution (CDC) schemes. We present an alternative approach which we call Collective-Drawdown CDC.
Assumes 6
- The market is complete, allowing any derivative on risk‑factors to be purchased. market
Source sentence
We also show that in com-plete markets, collective drawdown is optimal in the sense that under very mild assumptions on preferences, there are no mutually beneficial schemes that can improve upon collective drawdown.
- There is a finite number of members in each scheme. model
Source sentence
The key assumptions are: there are a finite number of members; at the time a member agrees to join the scheme, it is always in their best interests to do so; additional wealth 20 results in strictly better outcomes.
- The fund size is infinite for computational ease. execution
Source sentence
To begin with, we assume the fund size is infinite for computational ease.
- Mortality times are independently distributed and modeled with IFA’s S1PMA tables. data
Source sentence
We use the IFA’s S1PMA tables to model mortality, 13 assuming that the times of death of members are independently distributed.
Show 2 more
- All members contribute a fixed proportion of salary each year up to retirement. model
Source sentence
All members contribute a fixed proportion of their salary into the fund each year up to retirement.
- The economic model contains no risk factors other than those in the shared‑indexation CDC (i.e., constant interest rate, stock growth, volatility, wage growth and inflation). model
Source sentence
Our economic model is chosen for simplicity and so it contains no risk factors other than those shared-indexation CDC can be expected to face. We simulate our risky asset in a Black-Scholes market with constant interest rate r, stock growth µ and volatility σ.
Data
- IFA S1PMA mortality tables synthetic public
Source sentence
We use the IFA’s S1PMA tables to model mortality, 13 assuming that the times of death of members are independently distributed.
- Synthetic Black‑Scholes market simulation synthetic unknown
Source sentence
We simulate our risky asset in a Black-Scholes market with constant interest rate r, stock growth µ and volatility σ. Wage growth and inflation are also assumed constant.
Costs
Partial
Transaction costs occur at the investment layer and are reduced by netting of trades.
Source sentence
Transaction costs occur at the investment layer and are reduced by netting.
Finds 3
- All percentiles of the collective‑drawdown scheme outperform the standard dynamic‑accrual (shared‑indexation) scheme within five years. performance
Source sentence
Within 5 years, or less, all percentiles of the collective-drawdown scheme outperform the standard dynamic-accrual scheme.
- The median lifetime‑mean replacement ratio for collective‑drawdown CDC is 44%, compared to 35% for the standard dynamic‑accrual CDC. performance
Source sentence
The median lifetime-mean replacement ratio for collective-drawdown CDC is 44%, compared to 35% for standard dynamic-accrual CDC.
- In the collective‑drawdown scheme, all deciles of retirement income are increasing over time, whereas the median retirement income in the standard scheme is decreasing. performance
Source sentence
The median retirement income in the standard dynamic-accrual scheme is decreasing. In the collective-drawdown scheme, all deciles are increasing over time.
Checked by others
No other paper in the archive has been found to support or contradict these findings (compared with 8 related papers). Coverage is partial, so this is neither replication nor refutation.
1 paper in the archive
Code
No repository by the authors found in the paper.
Extracted 2026-11-14 · same card as JSON · A language model chose these items: check the source sentence before relying on one.
Cited References (50)
- Inter-generational crosssubsidies in the UK’s first CDC pension scheme
- Collective defined contribution ( CDC ) — link
- Fiscal risks and sustainability report – S eptember 2024 – charts and tables: supplementary tables (2024) — link
- The Equality Act (Age Exceptions for Pension Schemes) Order 2010 (SI 2010/2133) (2010) — link
- The Occupational Pension Schemes (Collective Money Purchase Schemes) Regulations 2022, UK Statutory Instruments 2022 No. 255 (2022) — link
- How might CDC develop in the UK ? (2024) — link
- Quantifying multi-employer and single employer CDC outcomes (2024)
- Collective Pensions with Investment Choice: Making CDC work for the UK (2025)
and 42 more
- Options on tontines: An innovative way of combining tontines and annuities (2019)
- On the optimal combination of annuities and tontines (2020)
- A new approach for satisfactory pensions with no guarantees (2020)
- Modern tontine with bequest: Innovation in pooled annuity products (2019)
- Bringing cost transparency to the life annuity market (2014)
- Optimal retirement income tontines (2015)
- EQUITABLE RETIREMENT INCOME TONTINES: MIXING COHORTS WITHOUT DISCRIMINATING (2016)
- Tontines with mixed cohorts (2021)
- Royal Mail Collective Pension Plan, Definitive Trust Deed and Rules (2024) — link
- The Collective Plan’s Handbook (2024) — link
- Pension Schemes Act 2021 (2021) — link
- The Case for Collective DC (2013) — link
- Stochastic modeling of assets and liabilities with mortality risk (2021)
- The pricing of options and corporate liabilities (1973)
- Martingales and arbitrage in multiperiod securities markets (1979)
- Martingales and stochastic integrals in the theory of continuous trading (1981)
- Stochastic finance: an introduction with examples (2023)
- King’s C omputational R esearch, E ngineering and T echnology E nvironment ( CREATE ) (2024) — link
- An analysis of the D utch-style pension plans proposed by UK policy-makers (2022)
- Pension Contracts and Risk Sharing–A Level Playing Field Comparison (2015)
- Intergenerational risk sharing and endogenous labour supply within funded pension schemes (2014)
- Intergenerational risk-sharing and risk-taking of a pension fund (2008)
- Intergenerational risk sharing within funded pension schemes (2011)
- The crisis in retirement planning (2014)
- Lifetime portfolio selection under uncertainty: The continuous-time case (1969)
- Optimal post-retirement investment and consumption under longevity risk in collective funds (2025) — link
- Intergenerational cross-subsidies in UK collective defined contribution ( CDC ) funds — → 2411.13565
- Optimal mutual insurance against systematic longevity risk — → 2410.07749
- Collective Defined Contribution Schemes Without Intergenerational Cross-Subsidies (2025) — link
- Risk aversion with many commodities (1974)
- Temporal resolution of uncertainty and dynamic choice theory (1978)
- Substitution, Risk Aversion, and the Temporal Behavior of Consumption and Asset Returns: A Theoretical Framework (1989) — link
- Optimal consumption and portfolio choice for pooled annuity funds (2008)
- Risks for the long run: A potential resolution of asset pricing puzzles (2004)
- Long-run risks and financial markets (2007)
- Explaining asset pricing puzzles associated with the 1987 market crash (2011)
- The levered equity risk premium and credit spreads: A unified framework (2009)
- King William’s Tontine: why the retirement annuity of the future should resemble its past (2015)
- How CDC pension levels compare with other types of schemes (2020) — link
- Modelling Collective Defined Contribution Schemes (2015)
- Life-cycle preferences revisited (2013)
- Modeling assets and liabilities of a Finnish pension insurance company: a VEqC approach (2005)
2 of 50 cited works are in the corpus; resolved via arXiv source bundles, Semantic Scholar and OpenAlex. Corpus-wide: 2596 linked papers.